Unbiased Learning-to-Rank with Biased Feedback¶
Joachims, T., Swaminathan, A., & Schnabel, T. (2017). Unbiased Learning-to-Rank with Biased Feedback. Proceedings of the Tenth ACM International Conference on Web Search and Data Mining, 781-789.
Cited by¶
1 citation across 1 artifact.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Salience-as-Significance
- So the conditional distribution of "appears at rank 1" given "is the best answer" is not what the user's inference needs — the user needs the reverse conditional, and the construction guarantees the forward one is non-trivial even when the reverse carries no information.
This sourceShows that the ranked slate is generated under a presentation/selection rule (position bias) decoupled from true relevance — clicks reflect what was surfaced and where, not whether a result is genuinely the best answer.
- So the conditional distribution of "appears at rank 1" given "is the best answer" is not what the user's inference needs — the user needs the reverse conditional, and the construction guarantees the forward one is non-trivial even when the reverse carries no information.
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Not recorded. The single citation of this work carries no recorded support check.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:35f99d504ab9 · see in the full table